An Image Edge Detection Method Based On The Intensity Change Of The Neighborhood Of A Vector Field

The image segmentation is to partition a picture into several segments according to its characteristics and the rules. It is one of the most important technologies in image processing, and it's the first operation of image auto-analysis. It has been used in many realms. Searching for the fit algorithm is the key of image segmentation and a little bit difficult. Therefore, to research the methods is very important not only in theory but also in practice.The main contribution of the dissertation is to study the basic theories of the vector fields of neighborhood variation, and then it is used to describe visual characteristics and to establish the relative system. The result proves it is effective.First of all, we give the definition of image segmentation, and introduce the general methods used in segmentation nowadays. On the base of the analysis, we build up the vector fields of neighborhood variation; put our emphasis on the description of the image edges and the domains of visual similar grey-level with the vector fields. After that, we find out the two methods used to detect the edges, the one with the biggest vector field, and the other with the perpendicular vector-pair of the biggest variation difference. The neighborhood size has got the great influence. We analyze it and find out that the influence mainly has two aspects: one is to expand to the edges and the other related to noise suppression. The PC-simulation shows that the techniques discussed above are effective.We also discuss the theory and techniques of description, recognition and detection of the image texture of line-strips. We choose variety of image data to carry on the related PC-simulation, and the results show that the theory and techniques are effective.